[amsat-bb] Re: General Telemetry Question (PITFALL!)
Jim White
jim at coloradosatellite.com
Wed Dec 11 11:03:52 PST 2013
I have been using the free program CurveExpert for years for plotting
and calculating calibration data. That is, for deriving the cal equation
from test data.
http://www.curveexpert.net/
It lets you choose linear or quadratic and if quadratic you can pick how
many terms you want. It lets you very quickly experiment with the form
of the formula so you can see how to get the best fit. It also will
suggest the highest exponential value for the formula.
I generally build my table of test data, ADC counts vs measured values
(V, I, temp, etc.), in Excel. Then I copy and paste that table into
Curve Expert, click one button to make it fit the curve and another to
show me the coefficients of the formula. I copy and past those back
into Excel next to the table. Then I add another column to the table
with values calculated using the formula and coefficients. That's the
double check to be sure the formula is correct.
Of course you can also see the standard deviation and other 'goodness'
values in the CurveExpert data. But I find I can tell by just looking
at the fit plot if I got the data wrong or my measurements weren't
precise enough.
I find CurveExpert much easier to use and much more flexible than
Excels' curve fitting functions.
Jim
On 12/11/2013 11:43 AM, Robert Bruninga wrote:
>> To follow up on Bob's comment. If you send the raw analog sensor
> data...
>> Change calibration values if found to be wrong after launch...
> We did on PCSAT!
>
> Caution to Satellite Builders: Be careful when using an EXCEL TREND LINE
> equation for doing Engineering Unit conversion back to original units. It
> was a big lesson for us back on PCSAT in 2001.
>
> The problem is, generally, EXCEL displays trend line equations in a nice
> GENERAL human readable form. For example, for our thermistors, the 3rd
> order trend line equation to convert from telemetry count back to degrees
> C was displayed by EXCEL as something like this: 2E-7 X^3 - 2E-4 X^2 +
> 1.804E-1 X + 2.379E2.
>
> One would think one is getting a very precise to 4-significant digit
> equation. WRONG. Notice the Cubed and Squared terms (which can be very
> big at warmer temperatures) are only represented to a single decimal
> digit(+/- 10%)!!! (2 and 2)...
>
> When this trend line is used (as displayed), to give back our
> temperatures from the incoming COUNT, the temperatures were way off!
>
> The key is to make sure the trend line equation is displayed in SCIENTIFIC
> format before you write it down and then try to use it. Then the first
> two terms above are properly displayed by EXCEL as 1.544E-7 X^3 and
> -2.069E-4 X^2. (instead of 2E-7 and 2E-4).
>
> We catch this error in a lot of student's work...
>
> Bob, WB4APR
More information about the AMSAT-BB
mailing list